November 20, 2026 · GMP
Invalidating OOS Results in 2026: Retesting, Human Error and What FDA and Health Canada Reject
By Mussarat Fatima

In November 2026 the United States Food and Drug Administration posted three warning letters that share one theme: laboratories that made out of specification results disappear. One active pharmaceutical ingredient maker invalidated more than 200 of them. A finished dose manufacturer approved a process validation report over a confirmed failure. A third firm released product with almost no testing at all. In the same week the European Compliance Academy released version 3.0 of its widely used out of specification guidance, its first full revision since 2013. Together these events send quality teams a single message. Inspectors are no longer satisfied by a retest that passes and a note that blames the analyst.
An out of specification, or OOS, result is a test result that falls outside the specification or acceptance criteria set in a marketing authorization, an official compendial monograph, or your own approved method. Invalidating that result, meaning formally setting it aside so it does not count against the batch, is one of the most scrutinised decisions a laboratory can make. Get it wrong and you do not just have a data integrity problem. You have a batch release problem, a recall exposure, and a warning letter waiting to happen.
This article explains when an OOS result can and cannot be invalidated under current FDA and Health Canada expectations, how the two phase investigation works, what retesting and resampling actually allow, and why human error and analyst retraining have become the fastest way to fail an inspection in 2026.
Executive Summary
- An OOS result can be invalidated only when a Phase I laboratory investigation finds a clear, documented and assignable cause, meaning a genuine laboratory error. A passing retest on its own is never enough.
- FDA's guidance Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022) and 21 CFR 211.192 remain the controlling references, alongside Health Canada GUI-0001 and sections C.02.011 to C.02.019 of the Food and Drug Regulations.
- A possible laboratory error is not enough to close an investigation at Phase I. When no conclusive laboratory cause is found, a Phase II full scale investigation into the manufacturing process must follow.
- Using human error and analyst retraining as a default root cause, without objective evidence, is now a leading warning letter finding.
- Averaging a failing result into passing numbers, using outlier tests to discard valid data, and testing into compliance through open ended retests are all rejected.
- In process validation an OOS result cannot be set aside simply because the batch was not distributed. A process that produces OOS batches is not in a state of control.
What does it mean to invalidate an OOS result?
Invalidating an OOS result means formally concluding that the result does not reflect the true quality of the batch, because a specific and demonstrated laboratory error caused it. Once invalidated, the result is excluded from the batch disposition decision. It matters because an invalidated OOS can make a failing batch look like a passing batch. For that reason regulators require objective evidence of the assignable cause, documented in a laboratory investigation, before any result is set aside.
The decision sits inside your deviation and investigation system. If you want the procedural backbone, see our guide to deviation and OOS investigations that survive a Health Canada inspection, and the companion piece on getting 21 CFR 211.192 right. This article focuses on the harder question that both touch on only briefly: the invalidation decision itself.
The two phase OOS investigation: Phase I and Phase II
The FDA framework splits every OOS investigation into two phases. Phase I is the laboratory investigation, which asks whether a laboratory error produced the result. Phase II is the full scale investigation into the manufacturing process, which is triggered whenever Phase I does not find a clear laboratory cause. The single most common failure is stopping at Phase I without evidence. The table below sets out how the two phases differ.
| Element | Phase I: Laboratory investigation | Phase II: Full scale investigation |
|---|---|---|
| Trigger | An OOS result is generated in the laboratory | Phase I finds no assignable laboratory cause, or the cause is inconclusive |
| Purpose | Determine whether a documented laboratory error caused the result | Investigate manufacturing, process, materials and sampling as the possible cause |
| Who leads | Analyst and laboratory supervisor | Quality unit, supported by production and engineering |
| Key actions | Review the method, calculations, standards, reagents, instrument and system suitability | Review the batch record, equipment, in process data, prior batches and retained samples |
| Retesting | Only to confirm or rule out a specific hypothesised laboratory error, under a pre approved plan | Resampling only when the original sample is shown to be unrepresentative |
| Outcome | Invalidate only with a confirmed, documented cause | Confirmed OOS stands; reject or make a documented risk decision; open CAPA |
When can you invalidate an OOS result? The conditions FDA and Health Canada accept
You can invalidate an OOS result only when three conditions are met together. First, a Phase I investigation identifies and documents a specific, assignable laboratory cause, such as a demonstrably diluted standard, a recorded instrument malfunction, or a confirmed calculation error. Second, that cause is scientifically sound and supported by objective evidence, not merely plausible. Third, the full record is retained, including the original failing result, which is explained but never deleted. If any of the three is missing, the result stands.
The point regulators now stress is that a possible cause is not a confirmed cause. In its 6 November 2026 warning letter to an active pharmaceutical ingredient maker, FDA stated plainly that a possible laboratory error is insufficient to close an investigation at Phase 1. A retest that happens to pass is not evidence that the first result was wrong. If you cannot name and prove the error, you have not earned the right to invalidate. For the broader pattern behind these findings, see our analysis of why CAPA keeps failing on root cause.
Retesting versus resampling: the rules
Retesting and resampling are not interchangeable, and inspectors expect you to know the difference. Retesting is a new analysis of the original retained sample. Resampling is the collection and testing of a new sample from the batch. Each is tightly constrained, and neither may be used to test a batch into compliance. The table below summarises the rules.
| Question | Retesting | Resampling |
|---|---|---|
| Definition | Re analysis of the original, retained sample | Testing of a new sample taken from the same batch |
| When allowed | To confirm or rule out a specific hypothesised laboratory error, under a written, pre approved plan | Only with documented evidence that the original sample was not representative or was prepared incorrectly |
| How many | Defined in advance in the SOP, never expanded until a result passes | Justified case by case, never open ended |
| What it cannot do | Overwrite a valid OOS by producing a passing number | Replace a Phase II manufacturing investigation |
Why human error and analyst retraining fail as a root cause
Human error and analyst retraining fail as a default root cause because they are usually a conclusion without evidence. When an investigation assigns every failing result to the analyst and closes every CAPA with retraining, a reviewer cannot tell a real laboratory error from a convenient one. That is precisely what FDA described in its November 2026 enforcement.
At the API site cited on 6 November 2026, FDA investigators reviewed 30 OOS investigation records and found a consistent pattern. Initial OOS results were invalidated based solely on passing results after reanalysis. Phase II investigations were either not conducted or improperly attributed failing results to analyst or procedural error, even where Phase I found no conclusive laboratory error. Analyst retraining was applied uniformly as the corrective action. FDA concluded that more than 300 OOS results since 2021 had been inadequately investigated, of which more than 200 were invalidated. The agency also noted that investigations into microbial counts over specification did not include identification of the organisms, which it called a fundamental part of any microbial excursion investigation. The full letter is linked in the sources below, and the same themes run through our review of the most common mistakes made during root cause analysis.
Averaging, outlier tests and the statistics trap
Statistics can support an OOS investigation, but they cannot be used to make a valid failing result vanish. Averaging a failing result with passing results can mask variability and hide an OOS, so FDA restricts when averaging is acceptable and requires that individual results still meet the specification where the specification applies to each unit. Outlier tests may describe a data set, but they cannot be used to discard a value simply because it fails, especially for a chemical assay. USP general chapter 1010 on analytical data provides legitimate statistical tools, yet none of them permit deleting valid data to reach a passing conclusion.
These shortcuts are data integrity findings in disguise. They show up next to audit trail gaps and are a reason data integrity still dominates enforcement, as we set out in why 60 to 80 percent of FDA warning letters still cite data integrity and in our piece on audit trail review as the new data integrity battleground.
OOS in process validation: why PPQ cannot pass over failures
An OOS result during process validation cannot be set aside on the grounds that the validation batch was never sold. The purpose of process performance qualification is to demonstrate that the commercial process consistently produces conforming product. A process that generates OOS results, in validation or in routine production, is by definition not yet in a state of control.
FDA made this concrete in its 2 November 2026 warning letter to a finished dose manufacturer, cited under 21 CFR 211.192 and 211.100(a). The firm found OOS assay results at multiple stability timepoints in both process validation and commercial batches, then concluded the validation results had no quality impact because those batches were not released. FDA rejected the logic, noting that the firm continued to release commercial batches after the validation batches, with the same root causes listed, and that the validation report was approved despite a confirmed OOS with no scientific justification recorded. The lesson for every validation master plan is simple. A failing PPQ batch is a signal to investigate the process, not a number to explain away. If you are rebuilding that programme, our team covers process validation and validation master plan remediation as part of MFLRC's validation services.
OOS, OOT and OOE: know the difference
OOS, OOT and OOE describe three different signals, and confusing them weakens an investigation. An OOS result is outside the specification or acceptance criteria. An out of trend, or OOT, result is still within specification but outside the expected statistical trend, such as a stability value that is drifting toward a limit. An out of expectation, or OOE, result is an atypical value that is within specification and trend but unexpected for the method or sample. The key interface is that an OOT stability result can predict a future OOS and must be investigated before it becomes one, which is exactly where many stability programmes fall short.
What FDA and Health Canada expect: the regulatory anchors
FDA and Health Canada expect the same core discipline: investigate every OOS thoroughly, decide batch disposition on evidence, and never release product you have not properly tested. For FDA the anchors are the May 2022 OOS guidance, 21 CFR 211.192 for the investigation of discrepancies and failures, 21 CFR 211.165 for release testing, and 21 CFR 211.194 for laboratory records. For Health Canada the anchors are GUI-0001 and sections C.02.011 to C.02.019 of the Food and Drug Regulations, which cover testing, records and the quality control department.
Release testing is where this all begins, and skipping it removes any possibility of a real OOS programme. In a warning letter dated 26 October 2026, issued after a section 704(a)(4) records request with no on site inspection, FDA cited an over the counter drug maker that released finished products without identity, assay, impurity or objectionable microorganism testing, relied on supplier certificates of analysis without verifying them, and referenced the Chinese Pharmacopeia instead of the United States Pharmacopeia. If you are not testing, there is nothing to find out of specification. We examine the testing side in batch release without real testing and the supplier side in when your contract lab's warning letter becomes your problem.
Compliance checklist: an OOS programme that holds up
- A written OOS SOP that defines Phase I and Phase II, timelines, roles, and the retest and resample plan before any retest is run.
- A firm rule that no original OOS result is ever deleted, overwritten or excluded without a documented, assignable cause.
- An objective evidence standard for any laboratory error claim, so that possible or likely is never accepted on its own.
- A mandatory Phase II trigger whenever Phase I is inconclusive.
- Controls on averaging and a prohibition on using outlier tests to reject valid assay data.
- A requirement to assess all affected batches on the market when a confirmed OOS is found late.
- Microbial identification built into every microbiological excursion investigation.
- Trending of invalidated OOS rates by analyst, method and product, reviewed by the quality unit.
Common mistakes
- Treating a passing retest as proof that the original result was wrong.
- Using analyst error or retraining as a default corrective action.
- Closing a Phase I investigation on a possible, unproven laboratory error.
- Averaging a failing result into compliance.
- Setting aside a process validation OOS because the batch was not distributed.
- Running retests until one passes, with no pre approved number.
- Failing to retain the original raw data behind an invalidated result.
Frequently asked questions
When can an OOS result be invalidated?
Only when a Phase I laboratory investigation identifies and documents a specific, assignable laboratory error that is supported by objective evidence. The original result is explained and retained, not deleted. A passing retest alone does not justify invalidation.
How many OOS retests are allowed?
There is no fixed universal number. The number of retests must be defined in advance in your SOP and tied to a specific hypothesis about a laboratory error. Open ended retesting, continued until a result passes, is considered testing into compliance and is not acceptable.
Is human error an acceptable root cause for an OOS result?
Only when a specific human error is identified and supported by evidence. Assigning human error by default, and closing the investigation with analyst retraining, is now one of the most frequently cited OOS deficiencies. If Phase I cannot confirm a laboratory cause, a Phase II manufacturing investigation is required.
What is the difference between OOS, OOT and OOE?
An OOS result is outside the specification. An out of trend, or OOT, result is within specification but outside the expected statistical trend. An out of expectation, or OOE, result is an atypical value within specification and trend. An OOT stability result often predicts a future OOS, so it must be investigated promptly.
Can you average results to overcome an OOS?
No. Averaging cannot be used to hide a failing individual result where the specification applies to each unit or where averaging would mask variability. FDA restricts averaging and expects the investigation to explain the failing result, not dilute it.
Does Health Canada follow the FDA OOS guidance?
Health Canada does not adopt the FDA guidance as its own, but its expectations in GUI-0001 and sections C.02.011 to C.02.019 of the Food and Drug Regulations align closely with the same principles: thorough investigation, evidence based disposition and full records. Sites that export to the United States should meet both.
What happens if a confirmed OOS is found after the product is on the market?
You must assess the risk to every affected batch still within expiry on the market, document that assessment, and take action that can include a recall. FDA has specifically criticised firms that delayed acknowledging the impact of a confirmed OOS on distributed product.
How MFLRC can help
If you are not confident that your invalidated OOS results would survive an inspector's review, that is exactly the gap MFLRC closes. We develop and rewrite OOS and OOT SOPs, deliver laboratory investigation and root cause training, run GMP audits and gap assessments, remediate CAPA programmes, and support responses to FDA warning letters and Health Canada observations. Our team has twenty years of quality assurance and regulatory experience across pharmaceuticals, natural health products, cannabis and medical devices, with a Health Canada security cleared and approved Quality Assurance Person leading the work.
Our OOS Programme Health Check is a fixed scope review of your last twelve months of invalidated results. We identify the invalidations an inspector would challenge, map them to FDA and Health Canada expectations, and deliver a prioritised remediation plan you can act on immediately.
Conclusion
The 2026 enforcement wave has made the standard for invalidating an OOS result unmistakable. A passing retest is not evidence. A plausible analyst error is not a root cause. A validation batch that fails is not excused by never being sold. Invalidation is a privilege you earn with a documented, assignable cause and lose the moment the record cannot support it. Build the discipline into your SOPs now, trend your invalidations honestly, and treat every unexplained failure as a question about your process, not just your people. That is how a laboratory turns an OOS result into a defensible decision instead of a warning letter.
Sources and references
- FDA, Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022)
- 21 CFR 211.192, Production record review
- 21 CFR 211.165, Testing and release for distribution
- Health Canada, Good manufacturing practices guide for drug products (GUI-0001)
- FDA Warning Letter, Umicore Argentina S.A. (6 November 2026)
- FDA Warning Letter, LGM Pharma Solutions LLC (2 November 2026)
- FDA Warning Letter, Guangdong Bangze Medicine Co., Ltd. (26 October 2026)
Downloadable Resource
OOS Investigation and Invalidation Checklist
A one page checklist that walks your laboratory through a defensible Phase I and Phase II OOS investigation, including the three conditions for invalidation and the retest and resample rules.
File: MFLRC-OOS-Investigation-Invalidation-Checklist.pdf
Share with others
Tags
